JPH04278779A - Computerized method and output generating device - Google Patents

Abstract

PURPOSE: To improve the outputting quality of a fully colored picture which contains a smaller number of colors as compared with the number of colors indicated by an input signal from a reproducing device. CONSTITUTION: The corrected picture element value of a color 1 is calculated from the input picture element of the color 1 and the sum of the diffusion error value of the color 1 and the outputted picture element value of the color 1 is selected out of possible values close to the corrected picture element value of the color 1. Then the quantization error of the color l is calculated from the difference between the corrected picture element value and the outputted picture element value of the color 1. By using the quantization error, an influence is exerted upon the next pigment element of the same picture element.

Description

[Detailed description of the invention]

0001

[Industrial Application Field] The present invention relates to electronic image processing.
Monitors that display a relatively small number of colors, especially using halftoning and error diffusion techniques where the quantization error of one color for a given pixel is used to correct the quantization of other colors for that pixel. The present invention relates to a system for improving the output quality of full-color images reproduced by a printer that prints a relatively small number of colors.

0002

BACKGROUND OF THE INVENTION In full-color images, each pixel is typically displayed as one of 16 million colors. However, conventional electronic displays and computer monitors, such as those used in the IBM PS/2 with VGA, have a
Only 56 colors can be displayed. Therefore, the problem is to obtain the best full-color image display on a monitor that can only display a small number of colors. Similarly, conventional printers are limited in the number of colors in a full-color image that can be reproduced. The present invention attempts to overcome this color reproduction problem by utilizing halftoning and error diffusion techniques.

Halftoning The term halftoning used here is
Receives an input image in which each pixel can incorporate any number of shades of gray or shades of color, and produces an output image in which each pixel can incorporate any one of a smaller number of shades of gray or colors. electronic processing. With proper selection of the output pixel values, the output image will appear to be a complete complement of shadows when viewed from a distance. For a good overview of halftoning techniques, see Digital Halftoning by R. Ulichney.
ng)” (Cambridge, MA., MIT P.
(Ress, 1987).

Error Diffusion Error diffusion is a generally well-known technique for halftoning images. Its first use was usually by Robert Floyd and Lewis.
Steinberg (Louis Steinberg)
``An Adaptive Algo-rit to Spatial Gray Scale'' announced by
hm for Spatial Gray S
Cale)”, pages 36-37 of the 1975 SID International Symposium Digest of Technical Papers. Various variations of error diffusion are described by R. Ulich.
It is also discussed in the above-mentioned paper ``Digital Halftoning'' by John C. Ney.

How to Process Error Diffusion Error diffusion begins with an error ei,j diffused at each pixel location equal to zero. Mathematically, this can be expressed as: ei,j =0 ∀i,j The input pixels are then processed in sequence following processing of the selected pixel at the initial or first pixel location where the spreading error is determined. The processing steps for each remaining pixel are as follows.

[Example 1] Step 1 The corrected pixel value mpi,j is the input pixel value ip
It is calculated as the sum of i, j and the diffusion error value at that pixel location. Mathematically, this is represented by the following formula: mpi,j =ipi,j +ei,j Step 2 The output pixel value opi,j is selected as one of the possible output values qi close to the modified pixel value. Mathematically, this is represented by the following formula: opi,j = Q(mpi,j) In the above formula, Q(χ) is the available output value q close to χ
Select one of i. Step 3 The quantization error δi,j is calculated as the difference between the corrected pixel value and the output pixel value. Mathematically, this is represented by the following formula: δi,j = mpi,j −opi,j Step 4 For unprocessed pixel locations, the diffusion error is incremented by an amount proportional to the quantization error at the pixel location. Mathematically, this is expressed by the following formula, assuming Σcr,s = ν. ei+r,j+s =ei+r,j+s +cr,s
δi,j Many of the differences between various error diffusion techniques are
This is due to the change in the selection of the diffusion coefficient cr,s used in step 4. For example, according to the paper by Freud and Steinberg cited above, the coefficient is a constant;
An example of the use of coefficients of random variables can also be found in US Pat. No. 4,654,721.

Applying Error Diffusion Independently to the Three Color Planes Error diffusion is particularly easy to apply when the display palette is orthogonal to red, green, and blue. An orthogonal palette in this state means the following. If a red tint r exists in the display palette, a green tint g exists in the display palette, and a blue tint b exists in the display palette, then the color triplet r, g, b
, is also present in the display palette. A color input image will be treated as three input images. Where

Similarly, a color output image would be viewed as three output images. Where

With the orthogonal palette, the three input images that make up the color input image are processed independently to produce three output images using error diffusion, and these three output images are combined to produce the color output image. Let's start composing images. That is, this is as follows.

This process is capable of producing high quality color images with accurate colors and has been proven and used. However, artifacts appear in the output image, the main one being the amount of texture that is visually apparent during image reproduction.

[0007]

SUMMARY OF THE INVENTION The present invention aims to improve this process by minimizing this large artifact.

[0008]

Means and Effects for Solving the Problems The present invention utilizes what is referred to as "combined color error diffusion" processing for input signals representing full-color images, and outputs that reproduce visually improved colors. Generate an image. Particularly suitable for use with orthogonal color palettes. Specifically, a signal representing the color of a pixel of a full-color image is input to a processing device such as a personal computer, and an output device that can generate only a small number of colors compared to the number of colors displayed by the input signal; For example, the final image is displayed on a monitor display or printed out on a printer. processing the input signal in such a way as to obtain output in a form that best reproduces or displays a full color image for the person viewing the output product, i.e. the image on a monitor or printed out by a printer; Error diffusion is applied to the input signal using an improved technique as follows. An initial pixel location is selected from which the diffusion error will be measured. The input pixel signals of the color image are then sequentially processed in the following steps, starting with the diffusion error for each color (eg, three colors) at each pixel location of zero, ie, the location represented by the following equation:

[Equation 1] Step 1 The corrected pixel value of color 1 is calculated by converting the pixel at the first or predetermined position (immediately after the initial pixel position) to the input pixel value of color 1 at that pixel position + the diffusion error value of color 1. Calculate as. Step 2: the output pixel value of color 1 at the predetermined position is selected from among those that can be close to the corrected pixel value of color 1;
The color 1 quantization error is calculated as the difference between the color 1 corrected pixel value and the color 1 output pixel value. Step 3 The corrected pixel value for color 2 is similarly calculated for the pixel at a given location as the input pixel value for color 2 at that pixel location + the diffusion error value for color 2. Step 4: The output pixel value of color 2 at a given location is chosen from among possible approximate sums of the corrected pixel value of color 2 multiplied by a constant and the quantization error of color 1. Step 5 The quantization error for color 2 is calculated similarly to the quantization error for color 1. That is, the error is color 2
is calculated as the difference between the corrected pixel value of color 2 and the output pixel value of color 2. Step 6 The corrected pixel value for color 3 is similarly calculated for the pixel at a given location as the input pixel value for color 3 at that pixel location + the diffusion error value for color 3. Step 7 The output pixel value of color 3 at a given location is approximately the sum of the corrected pixel value of color 3, the product of a constant and the quantization error of color 1, and the product of another constant and the quantization error of color 2. Selected from among the available. Step 8 The quantization error for color 3 is calculated similarly to the quantization error for color 1. Step 9 At the unprocessed pixel location, the color 1 diffusion error is incremented by the calculated sum of the color 1 quantization error for the given pixel location. The color 2 and color 3 diffusion errors are similarly incremented at each location using the calculated sum of each color 2 and color 3 quantization error at a given pixel location.

These step sequences are not believed to be definitive since many other step sequences still produce the same output. An output in which the result of combining these outputs represents a full-color input image that uses a small number of colors compared to the number of colors in the input image, but with minimal artifacts such as the amount of texture visually apparent in the output image. It becomes an image.

0010

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention employs a technique that may be referred to as "combined color error diffusion," and in the preferred embodiment described, the present invention employs a three-color system: green is color 1, red is color 2, and blue is color 2. Applies to systems with three color planes used as 3. The technique uses the quantization error of one color of a given pixel to correct the other colors of that pixel, thereby improving the performance of monitors that display a relatively small number of colors or printers that print a relatively small number of colors. The purpose is to improve the quality of full-color image output. More specifically, in image processing using error diffusion, a modified input signal is compared to a set of possible output signals and one of the output signals is selected. Typically the "nearest" output signal is chosen. This process in which an output close to the input is selected is called "
It is called ``quantization''. The difference between the input and the selected signal is called the "quantization error." In monochrome (one color with many shades) images it is fairly easy to determine the "closest" color, but in multicolor (e.g. many shades of red, green, blue) images the closest color is determined. Choosing (and how to choose) is not obvious. Today, quantization of color images is accomplished by processing each color component (usually red, green, and blue) independently. In contrast, the present invention uses the quantization error produced by one color component of a given pixel to influence the quantization of the next color component of that same pixel, so that the rendered color image The grain size does not seem to be very coarse. The present invention essentially involves specifying a new method for quantizing multicolor images represented by orthogonal or separable color image palettes, taking advantage of calculations made by error diffusion.

First, using the system and method of the present invention,
A signal representing the color of a pixel in a full-color image is input to a processing device such as a personal computer, and the final result is a monitor display, printer printout, etc. that consists of only a small number of colors compared to the number of colors represented by the input signal. , generates output. Therefore, the input signal is processed in such a way that the output is obtained in a format that best reproduces and represents a full-color image for the person viewing the output product, i.e., the image on a monitor or printed out on a printer. This is desirable. To do this, using error diffusion of the input signal, starting from the diffusion error of each color at the position of each pixel at zero, that is, the position expressed by the following equation, the input pixel signal of the color image is sequentially processed in the following steps: .

Step 1 The corrected pixel value for color 1 is calculated as the sum of the input pixel value for color 1 and the diffusion error value for color 1 at the first or predetermined pixel location of the image.

##EQU00003## The first or predetermined pixel location could be any location within the image based on the selection of a convenient starting point. The starting point is the initial pixel location, where the pixel's input signal is used as the basis for determining the spreading error for the remaining pixel locations. Since the pixel at the first pixel location in the upper right corner of the image is usually a convenient choice as a starting point, the first or predetermined location is typically the second, i.e., horizontally adjacent upper right corner. This is the corner position. Step 2 The output pixel value of color 1 is the possible output value of color 1 that is close to the corrected pixel value.

[Outer 5] selected as one of the

Step 3 The quantization error for color 1 is calculated as the difference between the corrected pixel value for color 1 and the output pixel value for color 1.

##EQU00005## The implementation of steps 1 through 3 is shown in FIG. In the figure, component Σ represents all devices capable of calculating sums and differences of signals. Whether the signals should be added or subtracted depends on the sign +
It is indicated by a symbol or -. The block named Quantization Qc1 represents the component that performs the quantization operation. Although this operation may be performed computationally, lookup tables are typically used in its implementation. Step 4 The corrected pixel value for color 2 is similarly calculated as shown below.

Step 5 The output pixel value of color 2 is the possible output value of color 2 that is close to the sum of the products of the corrected pixel value of color 2 and the constant and the quantization error of color 1,

[Outside 6] Selected as one of the.

##EQU00007## Step 6 The quantization error for color 2 is calculated as in the following equation, similar to the quantization error for color 1.

##EQU00008## The implementation of these steps 4 to 6 is shown in FIG. Step 7 The corrected pixel value for color 3 is similarly calculated as shown below.

[Equation 9] Step 8 The output pixel value of color 3 is the sum of the corrected pixel value of color 3, the product of a constant and the quantization error of color 1, and the product of another constant and the quantization error of color 2. Possible color 3 output values close to,

[Outside 7] Selected as one of the.

##EQU00001## Step 9 The quantization error for color 3 is calculated as in the following equation in the same way as the quantization error for color 1.

##EQU00001## The implementation of these steps 7 to 9 is shown in FIG. Step 10 At the unprocessed pixel location, the color 1 diffusion error is incremented by an amount calculated from the color 1 quantization error at this first pixel location:

##EQU00001## The implementation of this step is illustrated in FIG. As shown, the counters are used to generate indices r, s, and when applied to the coefficient store, the coefficients,

[Example 8] will occur. The index r,s is also added to the address i,j to form the indexed address i+r,j+s, and when applied to the error store, the error,

[Outer 9] cause The generated coefficients are quantization errors,

[Outside 1
0] and added to the error in the error storage to update the error,

[Outer 11] generate. Error updated after delay,

[Example 12] is inserted into the error storage, replacing the previous version. This operation is repeated for each index generated by the counter. Step 11 The diffusion error of color 2 is similarly incremented as follows.

##EQU00001## The implementation of this step is illustrated in FIG. This execution is shown in Figure 4.
is similar to the implementation of , but for each color 1 quantization error, color 1 coefficient storage, and color 1 error storage, color 2 quantization error, color 2 coefficient storage, and color 2 error storage device has been replaced. Step 12 The diffusion error for color 3 is similarly incremented as follows.

##EQU00001## The implementation of this step is illustrated in FIG. This execution is shown in Figure 4.
is similar to the implementation of , but for each color 1 quantization error, color 1 coefficient storage, and color 1 error storage, color 3 quantization error, color 3 coefficient storage, and color It replaces the error storage device of 3.

As mentioned above, in this embodiment, color 1 is green;
Color 2 is red and color 3 is blue. Their constant is 1
．． 0, i.e. α12=1.0, α13=1.0, α23
= 1.0, and the sum of those coefficients is 1.0, that is,
νc1=1.0, νc2=1.0, νc3=1.0. Please note these following selections. In step 2,
The green output value is chosen to minimize the difference between the green input value and the green output value. In step 5, the red output value is chosen to minimize the difference between the sum of the green and red input values and the sum of the green and red output values. Then, in step 8, the blue output value is chosen to minimize the difference between the sum of the green, red, and blue input values and the sum of the green, red, and blue output values.

The coefficients of the error diffusion process may be a set of constants or preferably the coefficients of a random variable. The execution of the latter process is shown in FIG. As shown, a random number generator is used to provide a random number signal to be multiplied by each constant from the constant storage. A constant store is connected to the index generation counter and the coefficient store in FIGS. 4, 5, and 6, and the product of the random number and the constant is input to the coefficient store, the output of which provides a multiplier for the spreading error, respectively.

[0014]

The present invention is a significant improvement over the performance of independent color error diffusion, since the generated output images have almost non-coarse grain characteristics while requiring almost no additional calculations. I understand. It can also be used to improve shaky color images when used to correct quantization errors in color planes that have already been processed to correct other color planes. Additionally, the present invention is useful in color imaging applications, particularly in heavily quantized displays (e.g., liquid crystal displays) and shallow frame buffers (e.g., IBM Model 8514).
It would be effective for A).

[Brief explanation of the drawing]

FIG. 1 is a schematic of a first color plane quantization system in accordance with the present invention;

FIG. 2 is a schematic of a second color plane quantization system in accordance with the present invention.

FIG. 3 is a schematic of a third color plane quantization system in accordance with the present invention;

FIG. 4 is a schematic of an incremental system of diffusion errors in a first color plane in accordance with the present invention;

FIG. 5 is a schematic of an incremental system of diffusion errors in a second color plane in accordance with the present invention;

FIG. 6 is a schematic of an incremental system of diffusion errors in a third color plane in accordance with the present invention;

FIG. 7 is a schematic of a system for generating coefficients for random error diffusion.

Claims (18)

[Claims] 1. A computer method for producing high quality output representative of a full color input image by sequentially processing signals representative of input pixels of the color image in a device that produces a relatively small number of possible output colors. providing a set of output signal sources corresponding to possible output pixel values of an output color; and receiving input signals representing at least two color components of the input pixel at each pixel location of the color image. , select an initial pixel location among the pixel locations, select an output pixel value among the possible output pixel values of each of the color elements according to the input signal at the initial pixel location, generating a signal indicative of the diffusion error of each of the color components at each pixel location subsequent to the initial pixel location; summing the received input signal and the diffusion error signal to generate a signal indicative of a respective modified pixel value; selecting an output pixel value among possible output pixel values for each of the elements, providing a signal to the output source indicating the selected output pixel value, and calculating a quantization error for each of the color elements; , determining the difference between a corrected pixel value of a color element and a selected output pixel value of said color element, and generating a signal indicative of a quantization error of each of said color elements at each said pixel location; At each pixel location, the product of each constant and the quantization error of each of the preceding color elements at each pixel location is added to the respective modified pixel value of each color element at each location; said signals representing respective modified pixel values are modified by quantization errors of preceding color components to influence the selection of each signal representing said selected output pixel values to be supplied to said output source. A computer method comprising the step of providing. 2. Further, a first pixel position following the initial pixel position
starting from a pixel location of , incrementing a diffusion error at an unprocessed pixel location, the incrementing step including one coefficient of a first set of coefficients and one of the first set of coefficients at the first pixel location. of each such pixel location by a quantity calculated from the product of the color component and the quantization error.
the quantization error of the second color element at the first pixel location and the quantization error of the second color element at the first pixel location incrementing the diffusion error of the second color element at each such pixel location by the amount of one coefficient of the third set of coefficients and the next color element at the first pixel location; incrementing the diffusion error of the next color element of each such pixel location by an amount calculated from the product of the quantization error and the first color element of each set of coefficients; 2. Incrementing the diffusion error of each subsequent color element at each such pixel location by an amount calculated from the product of each quantization error of each subsequent color element at such pixel location. computer method. 3. The method further includes the step of generating a value of the set of coefficients used for incrementing the diffusion error, the step of generating a random number, and each of the generated random number and a constant. 3. A computer method according to claim 2, comprising the steps of: calculating a product with a constant selected from a set respectively and generating an output signal indicative of the value of the product as the value of the set of coefficients. 4. The computer method of claim 3, wherein the set of coefficients is regenerated after processing each pixel. 5. The computer method of claim 1, further comprising the step of setting a diffusion error for each color component at each pixel location of zero prior to sequential processing of the signals representing input pixels of the color image. 6. Output generation for producing an improved quality output representative of a full color input image in an apparatus for producing a relatively small number of possible output colors by sequentially processing signals representing input pixels of a color image. An apparatus comprising: output means for providing a set of output signals corresponding to possible output pixel values of an output color; and input signals representing elements of at least two colors of the input pixels at each pixel location of the color image. means for receiving; and error diffusion means for providing a signal indicative of a diffusion error of each of the color components at each pixel location with respect to the color component at the selected initial pixel location; first summing means for summing the received input signal and the spreading error signal to generate a signal indicative of a respective modified pixel value, and in response to the signal indicative of each modified pixel value;
quantization means for selecting an output pixel value from among possible output pixel values for each of the color elements at each of said pixel locations and providing a signal indicative of said selected output pixel value to said output means; a second generating a signal indicative of the quantization error of each of the color elements at each of said pixel locations in response to the difference between the corrected pixel value of and the selected output pixel value of said color element;
means connected to said quantization means and responsive to said quantization error signal, said means for summing each of said quantization means before providing said signal indicative of a respective modified pixel value to said quantization means; At a pixel location, add the product of each constant and the quantization error of each of the preceding color elements at each pixel location to the respective quantized pixel value of each color element at each location, and display said addition. third summing means for supplying a signal to said quantization means. 7. The output generation device of claim 6, wherein the quantization means includes a look-up table. 8. The method further comprising means for setting the diffusion error of each of the color elements to each pixel location equal to zero before sequentially processing the signals representing input pixels of the color image.
The output generating device described in . 9. Further, a first pixel position following the initial pixel position
means for incrementing a diffusion error at an unprocessed pixel location in response to said selection of an output pixel value by said quantization means at a pixel location of said one coefficient of a first group of coefficients; by a quantity calculated from the product of the first color element and the quantization error at one pixel location;
incrementing the diffusion error of the first color element at each such pixel location, one coefficient of a second group of coefficients and the quantization of the second color element at the first pixel location; incrementing the diffusion error of said second color element at each such pixel location by an amount calculated from the product of an error, one coefficient of a third group of coefficients, and said first pixel location; incrementing the diffusion error of said next color element at each such pixel location by an amount calculated from the product of said next color element with said quantization error of one of the respective groups of coefficients; incrementing the diffusion error of said further color element at each such pixel position by an amount calculated from the product of a coefficient and each quantization error of each further color element at said first pixel position; , and means for activating said output means to provide an output signal from said set of output signals corresponding to a selected output pixel value of a color element at each pixel location. The output generating device described in . 10. Further, means for generating a value of the set of coefficients used in the error diffusion means, the means for generating a random number, and the means for generating a value from each set of the generated random numbers and constants, respectively. 10. The output generating device according to claim 9, further comprising: means for calculating a product with a selected constant and generating an output signal indicating the value of the product as a value of the set of coefficients. 11. The output generation device of claim 10, wherein the set of coefficients is regenerated after processing each pixel. 12. A computer method for producing high quality output representing a full color image in a device that produces a relatively small number of output colors by sequentially processing signals representing input pixels of a color image, the output providing an output signal source corresponding to possible output pixel values of a color; receiving input signals representative of at least two color components of an input pixel at each pixel location of the color image; selecting an initial pixel position, selecting an output pixel value among possible output pixel values for each of the color elements according to the input signal at the initial position, and generating a signal indicative of the selected output pixel value. calculating a modified pixel value for the color element at a first pixel location following the initial pixel location;
selecting an output pixel value for the first color element from among possible output pixel values that are close to the modified pixel value for the first color element;
calculating a quantization error of a color element as the difference between a modified pixel value of the first color element and a selected output pixel value of the first color element; and a modified pixel value of the second color element. selecting an output pixel value of a second color element from among possible output pixel values that is close to the sum of a first constant and the quantization error of the first color element; calculating the quantization error of the next color element as the difference between the corrected pixel value of the second color element and the output pixel value of the second color element; out of possible output pixel values that are close to the sum of the product of the constant and the quantization error of the first color component, and the product of the third constant and the quantization error of the second color component. selecting an output pixel value of the next color element;
calculating the quantization error of the next color element as the difference between the modified pixel value and the output pixel value of the next color element; with each modified pixel value calculated as a difference from the output pixel value and summing the product of a successive constant and a quantization error of each of said first color element and successive preceding color elements; selecting an output pixel value of a subsequent color element; and selecting one coefficient of a first set of coefficients at an unprocessed pixel location;
Increment the diffusion error of said first color element at each such pixel location by a quantity calculated from its product with said quantization error in the color element of one of the second set of coefficients. by a quantity calculated from the product of a coefficient and the quantization error of the second color element at the first pixel location;
incrementing the diffusion error in said second color element at each such pixel location and further incrementing the diffusion error in said next color element at said first pixel location by one coefficient of a third set of coefficients; incrementing the diffused error of the next color element at each such pixel location by a quantity calculated from the product with said quantization error, one of the respective sets of coefficients;
incrementing the diffusion error of the further successive color element at each such pixel location by an amount calculated from the product of the respective quantization errors of each further color element at the first pixel location; , processing the signal representing the input pixel at the remaining pixel locations by repeating the calculation, selection, and increment steps similar to the method for processing the input pixel signal at the first pixel location, and the output color at each pixel location. generating an output signal corresponding to selected output pixel values of the computer to reproduce a color image. 13. The method further includes the step of generating a value of the set of coefficients used for incrementing the diffusion error, the step of generating a random number, and the step of generating a random number and a constant value of the generated random number and each constant. 13. The computer method of claim 12, comprising the steps of: calculating a product of each coefficient with a constant selected from a set of coefficients, and generating an output signal indicative of the value of the product as the value of the set of coefficients. 14. The computer method of claim 13, wherein the set of coefficients is regenerated after processing each pixel. 15. A computer method further comprising the step of setting the diffusion error of each of the color components to each pixel location equal to zero prior to sequentially processing the signals representing input pixels of the color image. 16. An output generation apparatus for producing a high quality output representation of a full color input image, wherein the apparatus produces a relatively small number of possible output colors by sequentially processing signals representing input pixels of the color image. means for providing a set of output signals corresponding to possible output pixel values for an output color; and means for receiving input signals representing at least two color components of the input pixel at each pixel location of the color image; selecting an initial pixel location among the pixel locations; and selecting an output pixel value among possible output pixel values for each of the color elements according to the input signal at the initial location;
means for providing a signal representative of said selected output pixel value to said means for providing a series of corresponding output signals; at each pixel location subsequent to said initial pixel location, the color component of each input signal of the color component; means for processing said received input signal to produce a modified pixel value that is the sum of an input pixel value of said color component and a diffusion error of said color component; Means for selecting an output pixel value from among possible output pixel values of each color element at each pixel position using a quantization error that is a difference between the output pixel value and the first pixel value at a predetermined pixel position. selecting the output pixel value of the second color element at a predetermined position as one of the possible output pixel values close to the modified pixel value of the first color element; selecting a modified pixel value of a second color element as one of possible output pixel values that is close to the sum of a product of a first constant and a quantization error of said first color element; The output pixel value of the next color element in is the modified pixel value of the next color element,
one of the possible output pixel values that is close to the sum of the product of a second constant and the quantization error of the first color component and the product of a third constant and the quantization error of the second color component; and selecting the output pixel value of each subsequent color element at a given position as the sum of the corrected pixel value of each color element and the product of the respective constant and the quantization error of all preceding color elements. selecting the output pixel value as one of the possible output pixel values near the initial pixel position; means for incrementing a diffusion error at a pixel location, the diffusion error being calculated from the product of one coefficient of a first set of coefficients and the quantization error of the first color component at the first pixel location; incrementing the diffusion error of the first color component at each such pixel location by an amount of one coefficient of a second set of coefficients and the second color component at the first pixel location; incrementing the diffusion error of the second color element at each such pixel location by an amount calculated from the product of the color element with the quantization error; and one of the third set of coefficients. incrementing the diffusion error of the next color element at each such pixel location by an amount calculated from the product of the coefficient and the quantization error of the next color element at the first pixel location; each further color at each such pixel location by an amount calculated from the product of one coefficient of each set of and the respective quantization error of each further color element at said first pixel location. incrementing an element's diffusion error; and means for providing an output signal from the set of output signals corresponding to a selected output pixel value of a color element at each pixel location. an output generating device comprising: means for activating; 17. Further, means for generating values of the set of coefficients used in the error diffusion means, the means for generating random numbers, the means for generating random numbers, and the means for generating values of the set of coefficients used in the error diffusion means, respectively from the generated random numbers and the respective sets of constants. 17. The output generating device of claim 16, comprising means for calculating a product with a selected constant and generating an output signal representative of the value of the product as the value of the set of coefficients. 18. The output generation device of claim 17, wherein the set of coefficients is regenerated after processing each pixel.